基于混合Logit模型的高速公路交通事故严重程度分析

Severity of Traffic Crashes on Freeways Based on Mixed Logit Model

为从多方面掌握影响高速公路事故严重程度的因素,基于统计分析方法构建事故严重程度模型,分析其与道路、环境、驾驶员及车辆等因素间关系。鉴于多项Logit模型难以解析异质性及各因素对事故影响的交互作用,构建了混合Logit模型,并提出了刻画参数间相关性的方法。结果表明,考虑参数间相关性的混合Logit模型比多项Logit模型有更好的拟合优度,且能更合理地反映各因素对事故严重程度的作用效果;碰撞护栏或桥墩、女性驾驶员或驾驶员超过56岁时,更易产生受伤和死亡事故;能见度低于200m、驾驶员驾龄小于3年或超过10年、责任车辆为重型货车或车辆变更车道时,发生财产损失事故的概率增加,而发生死亡和受伤事故的概率有所降低;湿滑路面将导致受伤事故的概率增加3.7%,而混凝土护栏和夜间无照明时将使死亡事故的概率分别增加8.7%和28.8%。

In order to investigate factors affecting severity of traffic crashes,a model based on statistical analysis is developed,and relationships between severity of traffic crashes and influencing factors including road,environment,driver,and vehicle is analyzed.For a multinomial Logit model can barely capture unobserved heterogeneity and interactions among effects of factors on crashes,a mixed Logit model is developed and a method to analyze correlation between parameters is proposed.The results show that compared with the multinomial model,the mixed Logit model which considers correlation between parameters has a better goodness-of-fit and can more reasonably reflect effects of influencing factors on severity of traffic crashes.Colliding with barriers or piers,female drivers,or drivers over 56 are more likely to result in injuries and fatal crashes.Traffic crashes caused by visibility less than 200 m,driving experience of drivers is less than 3 years or over 10 years,and wrongdoers are heavy vehicles or lane-changing vehicles,are all associated with increasing probability of property-damage-only crashes,while with decreasing probability of injuries and fatal crashes.Wet and slippery road surfaces increases probability of injury crashes by 3.7%,while a concrete barriers and unlighted conditions at night result in increasing fatal crashes by 8.7%and 28.8%,respectively.